How accurate is the square-root-of-time rule in scaling tail risk: A global study

Jying-Nan Wang
a,
, Jin-Huei Yeh
b
, Nick Ying-Pin Cheng
c,d
a
Department of Finance, Minghsin University of Science and Technology, Taiwan
b
Department of Finance, National Central University, Taiwan
c
Department of Finance, Vanung University, Taiwan
d
Department of Finance, Yuan Ze University, Taiwan
a r t i c l e i n f o
Article history:
Received 9 March 2010
Accepted 18 September 2010
Available online 25 September 2010
JEL classication:
G18
G20
C20
Keywords:
Value at risk
Square-root-of-time rule
Jump diffusion
Serial dependence
Heavy-tail
Volatility clustering
Subsampling-based test
a b s t r a c t
The square-root-of-time rule (SRTR) is popular in assessing multi-period VaR; however, it makes several
unrealistic assumptions. We examine and reconcile different stylized factors in returns that contribute to
the SRTR scaling distortions. In complementing the use of the variance ratio test, we propose a new intu-
itive subsampling-based test for the overall validity of the SRTR. The results indicate that serial depen-
dence and heavy-tailedness may severely bias the applicability of SRTR, while jumps or volatility
clustering may be less relevant. To mitigate the rst-order effect from time dependence, we suggest a
simple modied-SRTR for scaling tail risks. By examining 47 markets globally, we nd the SRTR to be leni-
ent, in that it generally yields downward-biased 10-day and 30-day VaRs, particularly in Eastern Europe,
Central-South America, and the Asia Pacic. Nevertheless, accommodating the dependence correction is a
notable improvement over the traditional SRTR.
2010 Elsevier B.V. All rights reserved.
1. Introduction
Following several serious nancial crises in little more than a
decade, including the Asian Financial Crisis of 1997, the Dot-Com
Bubble of 2000, and the Global Financial Tsunami of 2008, risk
management, particularly in relation to tail risks, has recently in-
creased considerably in importance in numerous subelds of -
nance. Value at Risk (VaR), dened as a worst case scenario in
terms of losses on a typical day, is a popular measure of tail risk
management that is not only recommended by banking supervi-
sors (BCBS, 1996a), but is also widely used throughout the nancial
industry, including by banks and investment funds, see Prignon
and Smith (2010a,b). It is even used by nonnancial corporations
in supervising in-house nancial risks following the success of
the J.P. Morgan RiskMetrics system.
Operationally, tail risk such as VaR is generally assessed using a
1-day horizon, and short-horizon risk measures are converted to
longer horizons. A common rule of thumb, borrowed from the time
scaling of volatility, is the square-root-of-time rule (hereafter the
SRTR), according to which the time-aggregated nancial risk is
scaled by the square root of the length of the time interval, just
as in the BlackScholes formula where the T-period volatility is gi-
ven by r

T
_
. Regulators also advocate the routine use of the SRTR.
For example, to avoid duplication of risk measurement systems,
nancial institutions are allowed to derive their two-week VaR
measure by scaling up the daily VaR by SRTR; see, for example,
BCBS (1996b). In fact, horizons of up to a year are not uncommon;
many banks link trading volatility measurement to internal capital
allocation and risk-adjusted performance measurement schemes,
which rely on annual volatility estimates by scaling 1-day volatility
by

252
_
.
If the SRTR is to serve as a good approximation of all quantiles
and horizons, it not only requires the iid property of zero-mean re-
turns, but also that of the Normality of the returns. These pre-
assumptions are far from being realized in real world nancial as-
set returns, provided the numerous documented stylized facts that
are conict with these properties. Accordingly, numerous studies
have attempted to identify how these different effects give rise
to bias in SRTR approximation. The rst attempt is based on the
0378-4266/$ - see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankn.2010.09.028

Corresponding author. Tel.: +886 3 559 3142x1890; fax: +886 3 559

3142x3412.
E-mail addresses: jnwang@must.edu.tw (J.-N. Wang), jhyeh@ncu.edu.tw (J.-H.
Yeh), nickcheng@mail.vnu.edu.tw (N.Y.-P. Cheng).
Journal of Banking & Finance 35 (2011) 11581169
Contents lists available at ScienceDirect
Journal of Banking & Finance
j our nal homepage: www. el sevi er . com/ l ocat e/ j bf
fact that asset returns may be weakly dependent, both in levels and
higher moments. As illustrated in Jorion (2001), the SRTR tends to
understate long-term tail risk when the return follows a persistent
pattern, but tends to overstate the tail risk of temporally-aggre-
gated returns if it displays mean-reverting behavior. Similarly,
the presence of volatility clustering, as well-documented in the case
of most nancial assets since Engle (1982), Bollerslev et al. (1992),
Bollerslev et al. (1994), under the dynamic setup, has been demon-
strated using detailed examples of how the common practice of
converting 1-day volatility estimates to h-day estimates by SRTR
scaling is inappropriate and yields overestimates of the variability
of long-horizon volatility. On this, see Diebold et al. (1997) and
Mller et al. (1990).
Numerous extant studies have demonstrated that asset returns
exhibit heavy-tails (Fama, 1965; Jansen and de Vries, 1991; Pagan,
1996). Although allowing for dynamic dependence in the condi-
tional variance partially contributes to the leptokurtic nature, the
GARCH effect alone does not explain the excess kurtosis in nan-
cial asset returns. On the one hand, this motivates studies to em-
ploy their empirical GARCH modeling with student-t or
generalized error distributions to account for heavier tails. On
the other hand, researchers have turned to models that generate
price discontinuities to resolve the empirical regularity. Research-
ers have long realized that nancial time series exhibit certain
unusual and extreme violent movements, known as jumps and
modeled using jump diffusions developed by Merton (1976) that
create discontinuous sample paths. See Andersen et al. (2002),
Pan (2002), Eraker et al. (2003), Becker et al. (2009), Cmara
(2009) for recent evidence on the prevailing phenomena of jumps
in price processes. Nonetheless, how the underlying jumps inu-
ence the SRTR approximation of longer-term tail risks remained
unclear until the work of Danielsson and Zigrand (2006). They
intuitively and clearly show that SRTR tends to underestimate
the time-aggregated VaR and the downward bias deteriorates with
the time horizon owing to the existence of negative jumps. How-
ever, it remains unseen if in general price jumps are not conned
to downside extreme losses only, would the SRTR-induced down-
ward-bias move in the other direction instead or become
negligible?
Although we sound different alarms from distinct perspectives
by disclosing SRTR scaling as being inappropriate and misleading,
with documented upward biases for some effects and downward
biases for others, it is unclear after all whether the overall validity
of the SRTR is appropriate or not for practical risk implementation
given that all these effects coexist in a given asset. However, this
paper is not merely concerned with individual effects, such as a
weak dependence of returns, volatility scaling, price discontinu-
ities or leptokurticity, as is the case for the literature on the time
scaling performance of the SRTR. Instead, we are interested in
the interactions among these stylized facts on the scaling of tail
risks via the application of the SRTR. To our knowledge, no previ-
ous investigation has reconciled the quality of approximation in
time-aggregated tail risks using the SRTR under various confound-
ing factors.
This study lls this void by rst devising a general framework
for disentangling and separately estimating the sensitivity toward
each systematic risk factor. To examine the overall performance of
the SRTR approximation and characterize the potential bias, we de-
ne a bias function using a benchmark VaR based on averaging a
set of subsampled non-overlapping temporal aggregated VaRs.
Based on Monte Carlo experiments, this investigation demon-
strates that dependence at the return level is the dominant bias
factor. The SRTR leads to a systematic underestimation (overesti-
mation) of risk when the return follows a persistent (mean-revert-
ing) process, and can do so by a substantial margin. Moreover, the
magnitude of downward (upward) bias increases with the time
horizon. However, volatility clustering tends to drive the time-
aggregated VaR to slightly underestimate its true value. Alterna-
tively, the heavy-tailed nature of the underlying return overstates
the time-aggregated VaR via the SRTR. Perhaps surprisingly, unlike
the solely unilateral downside jumps specied by Danielsson and
Zigrand (2006) that indicate a severe underestimation bias, the
Monte Carlo allowing for both sided jumps with Poisson arrival
performed in this study suggests that there is a slight overestima-
tion when scaling with the SRTR.
In view of these results, proper tests for a preliminary verica-
tion of the applicability of the SRTR in practice are required. This
study rst recommends a new informal but informative subsam-
pling-based test, complementing the variance ratio test developed
by Lo and MacKinlay (1988),
1
for empirical studies. Moreover, it also
contributes to the literature by suggesting a simple modied-SRTR
that is robust to the time dependence-induced biases. By utilizing
47 markets included in the MSCI index, including both developed
and emerging markets, this study demonstrates that the SRTR
underestimates 10-day and 30-day VaRs by an average of approxi-
mately 5.7% and 13%, respectively. We also observe that the severity
of downward bias is greater for emerging markets in Eastern Europe,
Central and South America, and the Asia Pacic. For some developed
markets, even when the model assumptions are violated, the SRTR
scaling yields results that are correct on average, as shown in the
global investigation. This occurs because the underestimation result-
ing from the dynamic dependence structure is counterbalanced by
the overestimation resulting from the excess kurtosis and jumps.
Hence SRTR scaling can be appropriate in some cases. Although its
widespread use as a tool for approximate horizon conversion is
understandable, caution is, however, necessary. We believe that
the use of certain pretests as we proposed beforehand is important
and may illuminate the applicability of SRTR in the practical approx-
imation of tail risks. Our newly-proposed modied-SRTR approach is
shown to be effective in alleviating the bias attributable to the rst-
order effect from time dependence and the dependence correction is
a notable improvement over the traditional unadjusted raw SRTR.
The remainder of this paper is organized as follows. Section 2
formally denes the time-aggregated VaR and SRTR scaling. Sec-
tion 3 then performs algebraic analysis, in conjunction with Monte
Carlo simulations, to disentangle each isolated different stylized ef-
fect on the SRTR. This section also briey reviews the variance ratio
test devised by Lo and MacKinlay (1988). Section 4 introduces the
suggested variance ratio test and a newly-developed subsample-
based test for pretesting the applicability of the SRTR. More impor-
tantly, we introduce a new tail risk scaling rulethe Modied-
SRTR. Section 5 subsequently summarizes the global empirical
study based on data from 47 developed and emerging markets in-
cluded in the MSCI index. Finally, Section 6 presents the
conclusions.
2. Time-aggregated value at risk
The 1-day VaR, dened as VaR (1), measures the maximum pos-
sible loss over one trading day under a given condence level
100 (1 c). Supposing that the initial investment of the asset
is$1 and R is the random rate of return, then, the asset value at
the end of this trading day is v = 1 + R. Then, the one-day VaR,
VaR (1), under 100 (1 c) condence level is dened as
VaR(1) = infr[P[R 6 r[ > c: (1)
1
Finding that using SRTR to estimate Sharpe Ratios causes bias when returns
exhibit serial dependence, Lo (2002) suggests using the variance ratio test as a pretest.
Other related works include Huang (1985) and Ayadi and Pyun (1994), among many
others.
J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169 1159
Following the concern of 718 (Lxxvi) in the Basel II Accord, we de-
note the condence level as 99% in this paper.
The VaR (1) of an asset can simply be estimated through the
quantile function of the historical returns. Supposing that a se-
quence of T daily log prices of an asset p
t

T
t=0
is available, then
its daily returns are r
t

T
t=1
, where r
t
= p
t
p
t1
. By letting q() de-
note the quantile function, given P(r
t
6 q(0.01)) = 0.01, the value
of VaR (1) is dened as q(0.01). However, in practice, it is usually
hard to estimate the regulatory h-day VaR, VaR (h), since the time
horizon needed for the VaR (h) is quite long, especially when h is
large. For example, if we want to obtain the VaR (10) of an asset,
10 years of stock data may generate only 250 observations of 10-
day returns (250 trading days per year). Therefore, the Basel Com-
mittee on Banking Supervision suggests that banks scale VaR (1) up
to 10 days using SRTR. More generally, it says that VaR (1) could be
converted to VaR (h) by multiplying by

h
_
. Under the I.I.D. and
zero mean assumptions, without doubt,

h
_
VaR (1) is equal to
VaR (h).
The purpose of this paper is to investigate the validity of the
SRTR in time-aggregated VaR. Before proceeding with our algebraic
analysis, we need a true VaR (h) as a benchmark for the compari-
son. In practice it is usually difcult if not possible to estimate
the regulatory benchmark VaR. In this study, we recommend nd-
ing the benchmark h-period VaR, VaR (h), through a subsampling
scheme on the return series, and then use it for further character-
ization of the biases in our Monte Carlo experiments as well as in
empirical studies. This quantity is also employed to develop an
informative pretest for examining the overall applicability of the
SRTR in reality in a later section.
Before considering VaR (h), we need to generate the daily prices
where the sample size may not be too short and construct an h-day
return series from the original data. By leaving the rst h prices as
seeds, one may begin by subsampling the price series with a xed
length of h days with one of the seeds as the starting points. In this
regard, we confront a total of h 1 different subsamples of h-hori-
zon return series and the kth subsample time-aggregated return
from a non-overlapping interval, denoted by R
k
s
(h)
(T=h)1|
s=1
, is de-
ned as
p
hk
p
k1
|{z}
R
k
1
(h)
; p
2hk
p
hk1
|{z}
R
k
2
(h)
; . . . ; p
shk
p
(s1)hk
|{z}
R
k
s
(h)
; . . . ;
where k = 1, 2, . . . , h 1. Given P(R
k
s
(h) 6 Q
k
(0:01)) = 0:01, a specic
h-day VaR denoted by VaR
k
(h) can be computed intuitively by
Q
k
(0.01) for the kth series. For each k, there will be a corresponding
VaR
k
(h). As these VaR
k
(h)
h1
k=1
are obtained from different prices
without overlapping return periods, a benchmark naturally arises
to be dened as
VaR(h) =
1
h 1
X
h1
k=1
VaR
k
(h): (2)
Supposing the SRTR is correct, the scaled

h
_
VaR (1) shall be equal
to VaR (h). To examine whether the SRTR is tenable to serve as a
good approximation for the multi-horizon VaR, given our subsam-
pled and averaged benchmark VaR (h), we dene a bias function
f(h) to measure the approximation error of the SRTR in scaling tail
risks by,
2
f(h) =

h
_
VaR(1)
VaR(h)
1: (3)
If f(h) is positive (negative), using SRTR produces an overestimated
(underestimated) time-aggregated VaR. In the next section, we con-
struct the Monte Carlo explorations for different non-I.I.D. returns
features to investigate the inuences of serial dependence, heavy-
tailed distributions, jumps, and volatility clustering on the time-
aggregated VaR, respectively.
3. Characterizing biases: algebraic analysis with Monte Carlos
Different biases arise from different data generating processes,
except for the pre-assumed I.I.D. Gaussian case with zero mean.
To accommodate a wide spectrum of stylized facts documented
in the literature, we consider a fairly general data generation pro-
cess (DGP) for daily return r
t
that follows a non-zero mean ARMA
(1, 1)-GARCH (1, 1) model with Poisson jumps as
r
t
= l /
1
r
t1
a
t
h
1
a
t1
J
t
;
a
t
= r
t

t
;
r
2
t
= a
0
b
1
r
2
t1
a
1
a
2
t1
;
(4)
where t = 1, 2, . . . , T and J
t
is a compound Poisson process with jump
size distributed as N(0; r
2
j
) and constant jump intensity k. We allow
GARCH (1, 1) to govern the evolution of the conditional variance of
a
t
over time. {
t
} is a sequence of I.I.D. N (0, 1), a
0
> 0, a
1
P0, b
1
P0,
and a
1
+ b
1
< 1. Assuming there are 250 trading days per year, we let
a
0
=
0:15 (1 a
1
b
1
)
250
; (5)
simply to control the annualized volatility to be roughly about 15%.
Through the Monte Carlo simulation, a sequence of 5000 daily
returns, r
1
, r
2
, . . . , r
5000
, is constructed, which amounts to a sampling
period of about 20 years. Then we subsample (h 1) sequences of
h-day temporal aggregated returns from the above daily returns.
The VaR (1) of the daily returns is dened as the 1%-quantile for
this simulation. The time-aggregated VaR (h), h = 10 or 30, is com-
puted through the average of the subsampled quantities. With
2000 replications, we denote the true VaR (1) and VaR (h) as the
means of the 1-day and h-day VaRs, respectively. We examine
the following specic DGPs by restricting certain parameters to
isolate the different effects that might have an impact on the
time-aggregated VaR. We will come back and reconcile all these
impacts on the SRTR scaling of a tail risk later.
3.1. Non-zero mean
As the validity of SRTR scaling for quantiles hinges on a series of
assumptions, this subsection presents the bias arising from the pri-
mary factor of a non-zero mean for the underlying return. By let-
ting /
1
= h
1
= a
1
= b
1
= 0, r
t
= r and assume a zero jump (r
j
= 0), a
non-zero mean model is
r
t
= l a
t
; (6)
where a
t
= r
t
. Assuming Normality for simplicity, the daily VaR (1)
is
VaR(1) = l rU
1
(0:01); (7)
where U
1
() is the inverse normal cumulative distribution func-
tion. Straightforwardly, we can nd the VaR under a longer holding
period h to be
VaR(h) = h l

h
_
rU
1
(0:01): (8)
Therefore, the bias indicator f is
f(h) =
l rU
1
(0:01)

h
_
l rU
1
(0:01)
1: (9)
We let l = 0.08%, 0.04%, 0.02%, 0.02%, 0.04%, and 0.08%, which
imply that the means of their annualized returns are 20%, 10%, 5%,
2
Danielsson and Zigrand (2006) calculate VaR (h)/(

h
_
VaR (1)) as an indicator of
bias.
1160 J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169
5%, 10%, and 20%, respectively.
3
The simulation results are re-
ported in Panel A of Table 1. Interestingly, SRTR tends to underesti-
mate (overestimate) VaR (10) or VaR (30) due to the effect of a
negative (positive) l. However, as the readers will see in the follow-
ing subsection, we show that this non-zero-mean-induced bias is
slight and only of second-order importance in the presence of serial
dependence.
3.2. Series dependence
By xing l = a
1
= b
1
= 0, r
t
= r, and assuming a zero jump
(r
j
= 0), a weakly stationary ARMA (1, 1) model for r
t
is
r
t
= /
1
r
t1
a
t
h
1
a
t1
; (10)
where a
t
= r
t
.
The k-order autocorrelation of r
t
is
q
1
= /
1

h
1
r
2
var[r
t
[
; q
k
= /
1
q
k1
; for k > 1; (11)
where the variance of r
t
is
var[r
t
[ =
(1 2/
1
h
1
h
2
1
)r
2
1 /
2
1
: (12)
Then the variance of R
k
s
(h) is written as
var[R
k
t
(h)[ =
X
h1
i=0
X
h1
j=0
Cov(r
ti
; r
tj
) =var[r
t
[ h 2
X
h1
k=1
(h k)q
k
( )
= var[r
t
[ h 2
h
1 /
1

1 /
h
1
(1 /
1
)
2
!
/
1

h
1
(1 /
2
1
)
1 2/
1
h
1
h
2
1
! ( )
:
(13)
Therefore, based on the above result it is straightforward to show
VaR(h) =

var[R
k
t
(h)[
q
U
1
(0:01); (14)
where U
1
() is the quantile function from the standard Normal dis-
tribution. Then we can nd the bias f through (3), (12)(14), i.e.,
f(h) =

h
_
h2
h
1/
1

1/
h
1
(1/
1
)
2
!
/
1

h
1
(1/
2
1
)
12/
1
h
1
h
2
1
! ( )
1=2
1:
(15)
We let h
1
= 0 for the AR (1) models and consider /
1
= 0.7, 0.5, 0.2,
0.2, 0.5, 0.7. The MA (1) models, /
1
= 0 and h
1
= 0.7, 0.5, 0.2,
0.2, 0.5, 0.7 are also examined. The results in panel B and panel
C of Table 1 show that the SRTR yields severe overestimation (in the
case of negative serial correlation) or severe underestimation re-
sults (in the case of positive serial correlation) for VaR (10) and
VaR (30). According to (15), the patterns of bias function across
horizons, f(10), f(30), and f(90), are plotted in Figs. 1 and 2.
Table 1
Temporally-Aggregated VaR under Different Scenarios. We consider different DGPs of returns as shown in Eq. (4) where t = 1, 2, . . . , 5000. For each simulation, we generate the 10-
day and 30-day VaRs through the SRTR and the subsampling approach. We repeat 2000 times and report the means of

10
_
VaR (1), VaR (10), f(10),

30
_
VaR (1), VaR (30), and
f(30).

10
_
VaR (1)
VaR (10) f(10)%

30
_
VaR (1)
VaR (30) f(30)%
Panel A: Non-zero mean models with different l%
0.080 0.178 0.172 3.340 0.308 0.287 7.303
0.040 0.179 0.176 1.586 0.310 0.299 3.683
0.020 0.179 0.178 0.923 0.311 0.304 2.181
0.020 0.181 0.182 0.671 0.313 0.318 1.385
0.040 0.181 0.184 1.433 0.314 0.322 2.429
0.080 0.183 0.188 2.836 0.317 0.335 5.614
Panel B: AR (1) models with different /
1
0.700 0.252 0.513 50.876 0.437 0.986 55.714
0.500 0.208 0.335 37.969 0.360 0.606 40.610
0.200 0.184 0.220 16.534 0.318 0.385 17.394
0.200 0.184 0.153 20.114 0.319 0.261 21.917
0.500 0.208 0.128 62.641 0.360 0.212 70.004
0.700 0.252 0.119 111.642 0.437 0.190 129.847
Panel C: MA (1) models with different h
1
0.700 0.220 0.298 26.229 0.380 0.521 27.061
0.500 0.202 0.264 23.587 0.349 0.462 24.437
0.200 0.184 0.213 13.789 0.318 0.372 14.376
0.200 0.184 0.149 23.601 0.318 0.252 26.486
0.500 0.201 0.107 88.856 0.349 0.166 110.253
0.700 0.220 0.086 155.105 0.381 0.115 232.305
Panel D: GARCH (1, 1) models with different (a
1
, b
1
)
(0.130, 0.820) 0.193 0.198 2.746 0.333 0.338 1.431
(0.150, 0.800) 0.195 0.201 2.931 0.338 0.343 1.549
(0.130, 0.840) 0.198 0.203 2.526 0.342 0.349 1.987
(0.150, 0.820) 0.201 0.207 2.842 0.348 0.358 2.794
Panel E: Student-t models with different l
3.000 0.352 0.331 6.328 0.610 0.563 8.327
5.000 0.260 0.240 8.707 0.451 0.407 10.859
7.000 0.233 0.217 7.392 0.403 0.370 8.794
9.000 0.219 0.206 5.987 0.379 0.354 7.049
Panel F: Jump models with different (k, r
j
)
(0.058, 0.020) 0.185 0.184 0.511 0.320 0.317 1.006
(0.058, 0.030) 0.191 0.189 1.098 0.331 0.326 1.557
(0.082, 0.020) 0.187 0.185 0.992 0.324 0.319 1.499
(0.082, 0.030) 0.196 0.193 1.586 0.339 0.331 2.257
3
The average annual return of the examed 47 markets is 6.61%. Only two among
them, namely, Portugal and Turkey, have higher average annual returns that exceed
20%.
J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169 1161
3.3. Tail behavior
If r
t
is normally distributed, R
k
s
(h) is still normally distributed
due to the additive of the normal distribution. To obtain the 1%-
quantile of the normal distribution for the VaR calculation is not
hard. We can then calculate VaR (1) or VaR (h) using (14). Under
the Normal case, it is easy to nd that the bias function is only re-
lated to variances, i.e.,
f(h) =
h var[r
t
[
var[R
t
(h)[

1=2
: (16)
Nevertheless, without the Normal assumption, not only the vari-
ance, but also the tail behavior before/after temporal aggregation,
may affect f(h).
While the work of Dacorogna et al. (2001) has demonstrated
that, except for the boundary case of Normality, any other heavy-
tailed distribution under a stable law leads the SRTR to underesti-
mate the VaR, to our knowledge, there is no general theoretical
model that can provide a good interpretation of the relationship
between the time-aggregated VaR and tail behavior. In particular,
the question is a little involved from the distributional perspective
since there are some heavy-tailed distributions that may not be
closed under temporal aggregation, even if they are iid generated.
For instance, a student-t with 2 degrees of freedom would scale
like the SRTR in the tails; however, the fat-tail may no longer exist
after aggregation and thus this case gives rise to another source of
SRTR approximation error beyond the discussions in the literature.
In addition, while the leptokurticity of the observed returns may be
attributable to its underlying distributional property, it can also be
the consequence of a higher moment dependence such as volatility
clustering, as well as some occasional price discontinuities or
jumps. We will therefore investigate these issues through the
numerical analysis of three popular economic phenomena that
have contributed to the excess kurtosis in the stylized facts: vola-
tility clustering, heavy-tailed distributions and price jumps.
3.3.1. Volatility clustering
By letting l = /
1
= h
1
= 0 and assuming no jump, we consider
the following GARCH (1, 1) model
r
t
= r
t

t
;
r
2
t
= a
0
b
1
r
2
t1
a
1
r
2
t1
:
(17)
Drost and Nijman (1993) derive the temporal aggregation of the
GARCH processes and show, under regularity conditions, that the
corresponding sample path of R
t
(h) follows a similar GARCH (1, 1)
process with different parameters. The results have been suggested
to convert short-run volatility into long-run volatility in Christoffer-
sen and Diebold (1997) and Diebold et al. (1998). They do, however,
point out that using SRTR is inappropriate and produces overesti-
mates of the variability of long-horizon volatility. While these
works highlight the dangers of SRTR in the scaling of time-varying
volatilities into longer horizons, we take a different route. We con-
duct a series of Monte Carlo experiments to explore the robustness
of the SRTR in scaling VaR in the presence of GARCH effects in the
underlying return series.
We entertain pairs of (a
1
, b
1
) with a
1
+ b
1
~ 95 or 97 for the
GARCH (1, 1) models. Panel D of Table 1 shows that the SRTR tends
to yield only a slightly underestimated VaR (10) or VaR (30) in the
presence of volatility clustering in terms of the negative fs ranging
from 1.43% to 2.93%. These downward biases are intuitively
reasonable for overlooking the time-varying risks.
3.3.2. Heavy-tailed distribution
To demonstrate the effect of different distributional consider-
ations from the literature, we let the return process be
r
t
= rx
t
; (18)
where {x
t
} is a sequence of independent and identical student-t
distributions with m degrees of freedom. On the variance of the
aggregated entity, since {x
t
} is independent, var [R
t
(h)] is equal to
var [r
t
] multiplied by h. Nevertheless, it is intuitive that if we add
h daily returns to a h-day return, the long-tailedness appearing in
the daily return shortens as h increases. To see this, Fig. 3 shows
the probability density functions of daily and 10-day returns with
m = 5 and it is rather obvious that the tail part has been diluted after
aggregation.
By setting m = 3, 5, 7, and 9 to obtain different heavy-tailed re-
turn distributions from (18), we nd that an overestimated aggre-
gated VaR based on the SRTR is due to the heavier tail in the
daily return distribution. To probe further into what may have
gone wrong with the SRTR scaling approximation, we report both
the variance and kurtosis of the daily, 10-day and 30-day returns
under different DGPs in Table 2. By cross inspecting across the
0.7 0.5 0.3 0.1 0.1 0.3 0.5 0.7
0.6
0.2
0.2
0.6
1
1.4

h=10
h=30
h=90
Fig. 1. Bias functions of aggregated VaR with AR (1) models. Consider the model
from (10) given /
1
= 0. We plot different time horizons of bias functions, f(10),
f(30), and f(90), which are determined by (15).
0.7 0.5 0.3 0.1 0.1 0.3 0.5 0.7
0.5
0
0.5
1
1.5
2
2.5
3

h=10
h=30
h=90
Fig. 2. Bias functions of aggregated VaR with MA (1) models. Consider the model
from (10) given h
1
= 0. We plot the different time horizons of bias functions, f(10),
f(30), and f(90), which are determined by (15).
1162 J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169
Panel E in Table 1 and panel D in Table 2, it is readily seen that the
upward bias comes solely from the decrease in kurtosis, and not
from the scaling of volatility, since var[R
k
t
(10)] 10 var[r
t
]
whereas kurt[R
k
t
(10)] < kurt[r
t
]. Similar patterns can be found in
R
k
s
(30). Therefore, the SRTR overestimates VaR (10) and VaR (30)
by about 511% with different m for failing to take into account
the change in tail behavior among the original high frequency re-
turns and the temporally-aggregated ones in panel E of Table 1.
3.3.3. Jumps
While fat-tailed distributions may be more suited for daily inter-
nal risk management, they may not be suited for the modeling of
uncommon and unexpected systemic events. We consider a general
jump diffusion process
4
by letting l = /
0
= /
1
= h
1
= a
1
= b
1
= 0 and
r
t
= r to isolate the effect from jumps
r
t
= r
t
J
t
; (19)
where J
t
is a compound Poisson process with constant jump inten-
sity k and random jump size distributed as N(0; r
2
j
). The aggregated
variance of r
t
can be written as
var[R
t
(h)[ = var
X
h1
i=0
r
ti

X
h1
i=0
J
ti
" #
= h var[r
t
[: (20)
We let k = 0.058, 0.082 and r
j
= 2%, 3% to allow for a variety of com-
binations of jump intensities and sizes.
5
by cross inspecting across
the Panel F in Table 1 and panel E in Table 2, readers may nd the
results are similar to previous heavy-tailed cases: the problem with
the SRTR in jump models is not due to their variance scaling but to
their changing tail behavior. It is intuitive that a return distribution
with jumps has heavier tails. Furthermore, summing over h daily re-
turns smoothes out the infrequent jump effects. Thus, the SRTR pro-
vides an overestimated aggregated VaR.
We also nd var[R
k
t
(10)] 10 var[r
t
] while kurt[R
k
t
(10)] <
kurt[r
t
]. These simulation results shown in panel F of Table 1 indi-
cate that jumps indeed let the SRTR produce an overestimated
time-aggregated VaR. Nevertheless, under reasonable k and r
j
set-
tings, we also nd that the size of the systematic overestimation
bias from the SRTR only has a slight impact, as they are less than
1.6% in approximating VaR (10) and less than 2.3% in approximat-
ing VaR (30). Our result is largely different from that of Danielsson
and Zigrand (2006) who allow for downside jumps only in their
setup and it is reasonable to document downward bias in their
case. Instead, we not only nd upward bias via the SRTR scaling
in the presence of Poisson jumps, but, we document that the biases
may not be that much even after we consider some sizable jump
intensities and jump sizes.
3.4. A summary
The preceding analysis yields the following ndings. First, the
weak dependence in returns dominates among all the confounding
factors considered in this study when the SRTR is used to estimate
VaR (h). Positive serial dependence leads to a severe underestima-
tion in the SRTRs approximation of VaR (h), while severe overesti-
mation occurs in the case of negative serial dependence. Given
these results, this study proposes using the variance ratio test, typ-
ically employed to test for market efciency, to examine the syn-
thetic underlying serial dependence in empirical studies as a
pretest of the applicability of VaR scaling using the SRTR.
Second, using the SRTR may produce an overestimate or under-
estimate of VaR (h) because of the changes in tails. In cases of over-
estimates of the student-t distribution and jumps, the heavy-tailed
nature is smoothed out by aggregating daily returns. However, vol-
atility clustering may lead to the SRTR resulting in a slight down-
ward bias owing to neglecting the time-varying nature. To
summarize, this study carefully uses the SRTR to estimate the
time-aggregated VaR, when the real data exhibits serial depen-
dence, volatility clustering, a heavy-tailed distribution, or jumps.
10 8 6 4 2 0 2 4 6 8 10
0
0.1
0.2
0.3
0.4
0.5
Standardized 10day return density
Standardized 1day return density
Fig. 3. Probability densities of 1-day and 10-day returns. We generate the daily
returns which follow model (18) with m = 5. Then, we depict the probability
densities of standardized 1-day and 10-day returns.
Table 2
Variance and kurtosis of returns. We generate r
t
from the DGPs as (18) and (19). Thus
R
t
(10) can be calculated by r
t
. To characterize the potential biases attributable to the
scaling of variance or aggregation bias due to changing tail behavior, for each DGP we
report the corresponding variance and kurtosis of r
t
, R
t
(10), and R
t
(30).
Variance (%) Kurtosis
1-day 10-day 30-day 1-day 10-day 30-day
Panel A: AR (1) models with different /
1
0.700 0.117 4.881 18.148 2.998 2.991 2.970
0.500 0.080 2.080 6.876 2.998 2.979 2.952
0.200 0.063 0.899 2.776 3.000 2.987 2.958
-0.200 0.063 0.434 1.270 2.997 2.986 2.956
-0.500 0.080 0.303 0.838 3.003 2.990 2.964
-0.700 0.117 0.262 0.680 2.992 2.979 2.957
Panel B: MA (1) models with different h
1
0.700 0.089 1.648 5.090 3.002 2.990 2.968
0.500 0.075 1.288 3.985 2.998 2.996 2.975
0.200 0.062 0.841 2.576 2.999 2.980 2.966
-0.200 0.062 0.408 1.175 2.998 2.987 2.929
-0.500 0.075 0.209 0.508 2.998 2.975 2.945
-0.700 0.090 0.139 0.246 3.001 2.979 2.943
Panel C: GARCH (1, 1) models with different (a
1
, b
1
)
(0.130,0.820) 0.060 0.596 1.788 4.381 4.844 4.247
(0.150,0.800) 0.060 0.599 1.800 5.037 5.603 4.748
(0.130,0.840) 0.060 0.599 1.790 5.703 6.111 5.280
(0.150,0.820) 0.060 0.601 1.792 6.871 7.305 5.895
Panel D: Student-t models with different l
3.000 0.180 1.793 5.393 78.455 10.396 5.293
5.000 0.100 1.002 3.016 9.011 3.584 3.134
7.000 0.084 0.842 2.525 5.065 3.190 3.035
9.000 0.077 0.770 2.324 4.183 3.093 3.005
Panel E: Jump models with different (k, r
j
)
(0.058,0.020) 0.062 0.624 1.867 3.097 2.991 2.961
(0.058,0.030) 0.066 0.655 1.970 3.445 3.026 2.979
(0.082,0.020) 0.064 0.634 1.904 3.143 3.004 2.969
(0.082,0.030) 0.068 0.681 2.044 3.657 3.054 2.987
4
Danielsson and Zigrand (2006) demonstrate that the square-root-of-time rule
leads to systematic underestimation of risks, and their setup allows for downside
jumps that represent losses only. However, there is no a priori theoretical reason to
restrict, let alone expect, the prices to jump down only and therefore we entertain our
jump component to jump symmetrically and our parameters to be in line with the
jump diffusion literature.
5
Andersen et al. (2002) show that the jump intensity is about 0.014, that is, 14
times every thousand trading days on average, for the daily S&P 500 cash index. They
also estimate the jump size parameter r
j
to be at 1.5%. In the simulation conducted by
Huang and Tauchen (2005), their r
j
varies from 0 to 2.5%.
J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169 1163
4. Pretesting SRTR applicability and a modied-SRTR
4.1. Variance ratio test
From the viewpoint of empirical exploration, Lo and MacKinlay
(1988) propose a statistic to test the hypothesis of a random walk.
The general h-period variance ratio statistic VR (h) is denoted as
VR(h) =
var[R
t
(h)[
h var[r
t
[
= 1 2
X
h1
k=1
1
k
h

q
k
; (21)
where q
k
is the kth order autocorrelation coefcient of {r
t
}. When
VR (h) = 1, this means that {r
t
} follows the random walk hypothesis;
when VR (R) 1, {r
t
} exhibits serial dependence. Moreover, we re-
gard VR (h) as an indicator which measures the synthetical effects
on different degrees of serial dependence. If VR (h) is signicantly
larger (smaller) than one, we say this series is characterized by a
synthetically positive (negative) serial dependence of {r
t
}. It is intu-
itive that the positive (negative) serial dependence causes the SRTR
to be underestimated (overestimated). Lo and MacKinlay (1988) de-
ne the following statistic to estimate the VR (h) of Eq. (21),
VR(h) := 1 2
X
h1
k=1
1
k
q

^ q
k
; (22)
where ^ q
k
denotes the autocorrelation coefcient estimators. Under
the random walk hypothesis, VR(h) still approaches one. For the
standard inferences, it is necessary to compute its asymptotic vari-
ance. First denote a heteroskedasticity-consistent estimator of the
asymptotic variance of q
k
,
^
d
k
= T
X
T
j=k1
(r
j
^ l)
2
(r
jk
^ l)
2
" #
X
T
j=1
(r
j
^ l)
2
" #
2
(23)
where ^ l =
1
T
P
T
k=1
r
k
. Then, the following is a heteroskedasticity-con-
sistent estimator of the asymptotic variance of VR(h),
^
#
h
= 4
X
h1
k=1
1
k
h

2
^
d
k
: (24)
Regardless of the presence of general heteroskedasticity, the stan-
dardized statistic w*(h) can be used to test the hypothesis of a ran-
dom walk, i.e.,
w
+
(h) =

T
_
(VR(h) 1)

^
#
h
q -
a
N(0; 1); (25)
where -
a
denotes for asymptotically distributed as.
4.2. A new subsample-based test for overall SRTR applicability
While the variance ratio test is in spirit in line with the SRTR, it
offers only a partial picture since it is informative in detecting only
specically the dependence structure of the return series. As the
dependence structure is of rst-order importance, this paper offers
a new and simply complementary approach to test the overall
validity of applying the SRTR to scale VaR to a specic asset by
utilizing the subsamples we used to construct our benchmark
VaR (h). Since the way VaR
k
(h)
h1
k=1
is constructed is based on
R
k
s
(h)
(T=h)1|
s=1
, the h-period return from the non-overlapping sub-
sampling of the original prices, the subsampled VaR
k
(h)
h1
k=1
is em-
ployed to compute the benchmark VaR (h) by taking the subsample
average.
Before we arrive at a formal test for the SRTRs validity, a prop-
erly computed standard error of the bias term,

h
_
VaR(1) VaR(h),
is needed. Nonetheless, the well-documented time dependence
may carry over to the time-aggregated returns and thus the simple
variance estimator as dened by
R
0
(h) =
1
h 2
X
h1
k=1
VaR
k
(h) VaR(h) ( )
2
" #
1=2
;
may not be sufcient to accommodate the generality of the return
process. Moreover, the R
0
(h) dened above is also subject to the
small sample bias since h is commonly limited to 10 or 30.
To accommodate both of these concerns, we choose to rely on
the block bootstrapped samples to produce a reliable standard er-
ror for a formal test. This is done by taking into account the poten-
tial time dependence by retaining the dynamic structure of the
underlying returns by randomly drawing subsamples using blocks
of consecutive returns, thereby alleviating the small sample bias
problem and improving the testing performance. The implementa-
tion procedure is as follows.
We choose a block length of 10 and these blocks could be over-
lapping. Specically, the data are divided into T 9 blocks, with
the rst block being {r
1
, r
2
, . . . , r
10
}, the second block being
{r
2
, r
3
, . . . , r
11
}, . . . , . . . , and the last block being {r
T9
, r
T8
, . . . , r
T
}.
We then randomly resample T/10 blocks to construct a new boot-
strapped sample of T days of returns. For each bootstrapped resam-
ple, we calculate and save the values of VaR (1), the subsampled
and averaged VaR (h), and

h
_
VaR(1) VaR(h). By repeating the
above procedure for 5000 replications, we obtain a bootstrapped
sampling distribution of the bias,

h
_
VaR(1) VaR(h), based on
the 5000 bootstrapped resamples. Therefore, the bootstrapped
standard deviation of

h
_
VaR(1) VaR(h), dened as R(h), can
now be calculated.
Intuitively, under certain regularity conditions, it is easy to ar-
gue from the Central Limit Theorem
6
that

h
_
VaR(1) VaR(h)
R(h)
- N(0; 1): (26)
This statistic serves as our benchmark pretest for the overall valid-
ity of the SRTR in our subsequent analysis, after considering differ-
ent confounding dynamic and distributional properties that prevail
in real asset returns.
4.3. Scaling tail risk with a new modied-SRTR
As we have shown, the dynamic serial dependence in the return
process, among the other stylized features, serves as the rst-order
effect that biases the validity of the SRTR in scaling quantiles. In
view of this, the subsection moves one step further to propose a
simple and robust correction to the existing SRTR. It is a well-ac-
cepted fact that a variance ratio greater than 1 suggests the exis-
tence of a positive dependence in the underlying return series,
and the opposite situation holds true for the case of a VR (h) of less
than 1. This simple correction thus mainly makes use of the esti-
mated variance ratio as indicated in Eq. (22) to adjust the raw SRTR
by taking the time dependence structure into consideration.
7
Accordingly, we formally dene an estimator which estimates VaR
(h) through this robust rule as
MVaR(h) =

h VR(h)
q
VaR(1): (27)
We thus refer to it as the modied-SRTR (MVaR). Note that if a time
series is serially uncorrelated, the variance ratio is 1 and therefore
MVaR (h) will simply reduce to

h
_
VaR (1), which is essentially
6
The Central Limit Theorem holds for weakly dependent data as long it is
appropriately standardized by a standard error that accounts for the underlying
dependence.
7
We thank the anonymous referee for pointing this suggestion out to us.
1164 J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169
Table 3
Global Evidence on SRTR Scaling of Tail Risk. We consider 47 markets globally with both developed markets and emerging markets listed in the MSCI. The sample period extends
from January 2, 1997 to December 31, 2009 (3391 trading days). For each market, the time-aggregated VaR (10 days and 30 days) computed through both the simple SRTR and
subsampling approach are reported. We also report the overall bias, f, along with its signicance based on the block-bootstrapped standard error, as well as the statistic of the
variance ratio test, w*.

10
_
VaR (1)
VaR (10) f(10)% w*(10)

30
_
VaR (1)
VaR (30) f(30)% w*(30)
Panel A: Africa
Morocco 7.72 9.27 16.76 4.60
a
13.37 14.26 6.27 4.80
a
South Africa 12.16 13.27 8.32 1.50 21.07 28.25 25.43
a
1.90
c
Turkey 24.97 23.02 8.46 0.68 43.25 42.91 0.79 1.58
Average 14.95 15.19 5.54 2.26 25.89 28.47 10.30 2.76
Panel B: Europe
Austria 14.60 15.63 6.60 0.45 25.28 34.04 25.73
a
1.37
Belgium 10.80 12.55 13.93
c
1.26 18.71 23.31 19.73
b
1.21
Denmark 11.82 12.05 1.96 0.11 20.47 23.44 12.68 0.59
Finland 17.78 18.40 3.40 0.80 30.79 31.04 0.81 0.29
France 13.76 13.85 0.59 2.20
b
23.84 22.62 5.38 1.26
Germany 15.71 15.46 1.63 1.16 27.21 26.84 1.39 0.57
Greece 16.46 17.02 3.34 2.99
a
28.50 28.78 0.97 3.05
a
Ireland 13.61 16.53 17.69
a
0.93 23.57 30.27 22.13
a
1.26
Italy 12.58 14.14 11.05 0.81 21.79 24.83 12.27 1.15
Luxembourg 14.75 16.16 8.70 2.40
b
25.55 36.37 29.75
a
3.29
a
Netherlands 13.38 14.66 8.73 0.14 23.17 29.19 20.62
a
0.84
Norway 15.90 15.45 2.93 0.31 27.54 31.56 12.74 0.55
Portugal 10.40 11.34 8.22 2.38
b
18.02 22.00 18.09
c
3.04
a
Spain 12.91 13.29 2.86 0.99 22.37 22.74 1.64 0.14
Sweden 14.02 13.27 5.63 2.17
b
24.29 24.02 1.09 0.87
Switzerland 11.55 12.20 5.37 0.39 20.00 21.16 5.51 0.25
UK 11.07 10.99 0.78 1.64 19.18 19.56 1.95 1.00
Average 13.59 14.29 4.79 0.09 23.55 26.58 10.40 0.79
Panel C: Eastern Europe
Czech R. 13.23 15.70 15.77
b
1.06 22.91 31.81 28.00
a
1.63
Hungary 17.81 18.59 4.18 0.24 30.85 44.36 30.46
a
1.16
Poland 13.61 14.59 6.75 1.80 23.57 29.14 19.09
b
2.23
b
Russia 28.21 31.16 9.47 0.80 48.86 55.44 11.88 0.47
Average 18.21 20.01 9.04 0.97 31.55 40.19 22.36 1.38
Panel D: Central and South America
Argentina 21.55 21.72 0.77 1.50 37.33 43.24 13.67
b
1.82
c
Brazil 20.03 22.18 9.66 1.28 34.70 38.19 9.14 0.16
Chile 9.57 11.70 18.20
b
4.28
a
16.58 23.49 29.41
a
2.89
a
Mexico 14.18 14.29 0.80 0.60 24.56 27.01 9.07 0.69
Peru 15.63 18.27 14.44
c
2.86
a
27.07 38.48 29.65
a
3.15
a
Venezuela 18.08 22.65 20.15
a
0.63 31.32 47.28 33.75
a
1.31
Average 16.51 18.47 10.67 1.43 28.59 36.28 20.78 1.62
Panel E: North America
Canada 11.08 11.10 0.14 0.87 19.20 22.26 13.75 0.06
US 11.29 10.98 2.82 2.40
a
19.55 21.38 8.55 1.32
Average 11.19 11.04 1.34 1.64 19.37 21.82 11.15 0.69
Panel F: Asia Pacic
Australia 9.31 9.22 1.02 0.72 16.12 17.75 9.17 0.22
Hong Kong 16.02 16.38 2.23 0.53 27.74 31.27 11.27 0.13
India 15.49 15.77 1.78 0.42 26.82 28.16 4.73 0.95
Israel 10.71 10.86 1.35 0.81 18.55 18.90 1.85 1.88
c
Japan 13.75 11.30 21.71
b
1.71
c
23.82 24.88 4.25 1.03
Jordan 11.55 11.11 3.92 0.44 20.00 20.81 3.86 1.24
Korea 18.84 16.65 13.13
c
0.21 32.63 31.31 4.23 0.71
Malaysia 12.42 15.90 21.92
a
0.44 21.51 28.58 24.75
a
0.97
New Zealand 7.15 8.49 15.73 0.45 12.39 15.49 19.98 1.01
Pakistan 16.41 22.56 27.26
a
3.82
a
28.42 44.73 36.46
a
3.82
a
Philippines 14.56 14.70 0.96 2.44
b
25.21 28.84 12.59 3.29
a
Shanghai 16.51 15.45 6.81 0.52 28.59 24.88 14.90 1.53
Singapore 12.27 16.94 27.56
a
2.10
b
21.25 29.44 27.82
a
2.33
b
Thailand 16.01 19.36 17.30
a
2.04
b
27.73 35.59 22.08
a
2.93
a
Taiwan 13.74 14.09 2.54 1.12 23.79 25.67 7.32 1.93
c
Average 13.65 14.59 4.80 0.76 23.64 27.09 11.13 1.43
Total average 14.36 15.32 5.70 0.62 24.87 29.05 13.00 1.21
a
Those test statistics that are statistically signicant at the condence levels of 99%.
b
Those test statistics that are statistically signicant at the condence levels of 95%.
c
Those test statistics that are statistically signicant at the condence levels of 90%.
J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169 1165
the case for the typical raw SRTR. However, it should be noted that
the new scaling rule corrects for serial dependence only and thus
there remain other potential bias factors that might distort the scal-
ing of multi-horizon tail risks.
5. Empirical evidence
5.1. Data
To examine whether the SRTR is an appropriate method for
obtaining the time-aggregated VaR, this study employs daily stock
market index returns from 47 markets listed in the Morgan Stanley
Capital International (MSCI) index from Datastream, which in-
cludes most developed and emerging markets. The sample period
ranges from January 2, 1997 to December 31, 2009 (3391 trading
days). The markets are divided into six regions, Africa, Europe,
Eastern Europe, Central and South America, North America, and
the Asia Pacic. For each market, VaR (1) is obtained from the
1%-quantile of daily returns and the corresponding VaR (10) and
VaR (30) are calculated by (2). This study also reports the com-
puted biases from (3). Because the degree of series dependence
dominates in the SRTR approximation, we also examine its exis-
tence via the variance ratio tests for each market. Furthermore, this
study applies the newly-proposed test statistic in (26) to test
whether the overall SRTR induced biases are signicantly overesti-
mated or underestimated. The test statistics that are statistically
signicant at condence levels of 99%, 95% and 90% are marked
with ***,** and *, respectively in Table 3.
5.2. Main ndings
Table 3 lists the empirical results for each market. The means of
VaR (10) and VaR (30) are 15.32% and 29.05%, respectively. In pre-
senting this table in visual form, f(10) denotes the approximation
bias obtained using the SRTR to estimate VaR (10), which generally
yields downward bias of about 5.7%. Usually, severe downward
biases are associated with positive serial dependence, as indicated
by a positive and signicant w* statistic. Meanwhile, the bias grows
rapidly with an increasing horizon. For instance, when the time
horizon is increased to 30 days, the averaged understated bias
grows to 13%. Fig. 4 illustrates the results of Table 3 graphically,
and sketches the scatter plot of the benchmarks VaR (10) and
VaR (30) with their corresponding biases, f(10) and f(30), in per-
centage terms for each market. Except for Japan and Korea for
the 10-day horizon, the upward biases f are generally below 10%,
and are insignicant among the other nine markets experiencing
SRTR overestimation. However, the markets experiencing SRTR
underestimation of over 10% include 13 markets for VaR (10) and
26 markets for VaR (30).
To be specic, according to the subsample-based test statistic
presented in (26), while the SRTR-scaled 10-day VaR signicantly
underestimates the benchmark VaR (10) in 10 markets, usually
due to persistent returns, two markets (namely, Japan and Korea)
experience signicant SRTR overestimation that may be attribut-
able to their different mean-reverting behavior in terms of return
dependence. When considering a 30-day horizon, 18 markets are
signicantly underestimated, and none are signicantly overesti-
mated. As a whole, the above preliminary results suggest that SRTR
is a lenient rule for scaling longer-termtail risks, corresponding to a
situation of insufcient prudence and nancial institutions facing a
combination of extreme risk and inadequate capital requirements.
Most notably, the results differ among the six geographical
areas surveyed. Interestingly, North America displays the lowest
average tail risks. Meanwhile, while Eastern Europe and Central
and South America have larger VaRs than the other areas, their
average bias is roughly double that of the other areas. Although
the average SRTR bias, f(10), in the Asia Pacic is just 4.8%,
removing the overestimated outliers of Japan and Korea turns the
average f(10) in the Asia Pacic into 8.22%, equaling Eastern Eur-
ope and Central and South America. Within the Asia Pacic, there
are ve markets, namely, Malaysia, New Zealand, Pakistan, Singa-
pore, and Thailand, with the VaRs being underestimated by over
15%.
0 10 20 30 40 50 60
40
30
20
10
0
10
20
30
40
VaR (h) %

(
h
)

%
h=10
h=30
Fig. 4. Scatter Plot of VaR (h) with the corresponding f(h). Based on 47 markets, we
sketch two scatter plots of VaR (10) and VaR (30) with their corresponding f(10)
and f(30), respectively, from January 2, 1997 to December 31, 2009.
0 20 40 60 80 100
0
20
40
60
80
100
k
u
r
t
[
R
kt
(
1
0
)
]
0 20 40 60 80 100
0
20
40
60
80
100
kurt [r
t
] kurt [r
t
]
k
u
r
t
[
R
kt
(
3
0
)
]
Fig. 5. Scatter Plot of kurt[r
t
] with its corresponding scaled kurt[R
k
t
(h)]. Using 47 markets, we sketch two scatter plots of kurt[r
t
] with their corresponding kurt[R
k
t
(10)] and
kurt[R
k
t
(30)] from January 2, 1997 to December 31, 2009.
1166 J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169
Table 4
Global Evidence of Modied-SRTR Scaling. We recompute the time-aggregated VaR (10 days and 30 days) using the newly-proposed modied-SRTR scaling rule, MVaR (h), for the
47 countries regions and the overall bias f, along with its signicance based on the block-bootstrapped standard error, and the statistic of variance ratio test, w*.
MVaR (h) VaR (10) f(10)% w*(10) MVaR (h) VaR (30) f(30)% w*(30)
Panel A: Africa
Morocco 9.53 9.27 2.76 4.60
a
18.01 14.26 26.26 4.80
a
South Africa 13.04 13.27 1.69 1.50 24.06 28.25 14.83
c
1.90
c
Turkey 25.72 23.02 11.71 0.68 47.92 42.91 11.69 1.58
Average 16.10 15.19 4.26 2.26 30.00 28.47 7.71 2.76
Panel B: Europe
Austria 15.00 15.63 4.00 0.45 28.88 34.04 15.17
b
1.37
Belgium 11.56 12.55 7.92 1.26 20.81 23.31 10.72 1.21
Denmark 11.88 12.05 1.44 0.11 21.53 23.44 8.16 0.59
Finland 17.21 18.40 6.50 0.80 31.42 31.04 1.23 0.29
France 12.29 13.85 11.25 2.20
b
21.21 22.62 6.24 1.26
Germany 14.88 15.46 3.71 1.16 25.96 26.84 3.27 0.57
Greece 18.44 17.02 8.30 2.99
a
34.26 28.78 19.05
b
3.05
a
Ireland 14.29 16.53 13.54
b
0.93 26.37 30.27 12.88 1.26
Italy 13.06 14.14 7.64 0.81 23.82 24.83 4.08 1.15
Luxembourg 16.30 16.16 0.89 2.40
b
31.21 36.37 14.21 3.29
a
Netherlands 13.28 14.66 9.39 0.14 24.86 29.19 14.83
b
0.84
Norway 15.63 15.45 1.17 0.31 29.00 31.56 8.12 0.55
Portugal 11.65 11.34 2.76 2.38
b
22.55 22.00 2.51 3.04
a
Spain 12.29 13.29 7.53 0.99 22.64 22.74 0.45 0.14
Sweden 12.72 13.27 4.14 2.17
b
22.70 24.02 5.53 0.87
Switzerland 11.31 12.20 7.33 0.39 20.45 21.16 3.40 0.25
UK 10.09 10.99 8.13 1.64 17.33 19.56 11.40 1.00
Average 13.64 14.29 4.67 0.09 25.00 26.58 5.63 0.79
Panel C: Eastern Europe
Czech R. 14.11 15.70 10.11 1.06 26.78 31.81 15.81
b
1.63
Hungary 18.03 18.59 2.97 0.24 33.98 44.36 23.39
a
1.16
Poland 14.57 14.59 0.14 1.80 27.01 29.14 7.28 2.23
b
Russia 29.56 31.16 5.14 0.80 51.19 55.44 7.67 0.47
Average 19.07 20.01 4.59 0.97 34.74 40.19 13.54 1.38
Panel D: Central and South America
Argentina 22.94 21.72 5.61 1.50 42.20 43.24 2.39 1.82
c
Brazil 18.62 22.18 16.04
b
1.28 34.19 38.19 10.46 0.16
Chile 11.41 11.70 2.48 4.28
a
20.27 23.49 13.71 2.89
a
Mexico 14.55 14.29 1.80 0.60 25.82 27.01 4.38 0.69
Peru 18.51 18.27 1.32 2.86
a
35.57 38.48 7.55 3.15
a
Venezuela 18.63 22.65 17.72
b
0.63 34.37 47.28 27.30
a
1.31
Average 17.44 18.47 4.59 1.43 32.07 36.28 10.97 1.62
Panel E: North America
Canada 10.53 11.10 5.13 0.87 19.08 22.26 14.29 0.06
US 9.72 10.98 11.42 2.40
a
16.84 21.38 21.23
b
1.32
Average 10.13 11.04 8.28 1.64 17.96 21.82 17.76 0.69
Panel F: Asia Pacic
Australia 8.92 9.22 3.16 0.72 15.78 17.75 11.13 0.22
Hong Kong 15.51 16.38 5.32 0.53 28.08 31.27 10.21 0.13
India 15.75 15.77 0.10 0.42 28.60 28.16 1.57 0.95
Israel 11.06 10.86 1.83 0.81 20.91 18.90 10.64 1.88
c
Japan 12.48 11.30 10.44 1.71
c
21.50 24.88 13.59 1.03
Jordan 11.99 11.11 7.87 0.44 22.59 20.81 8.59 1.24
Korea 18.67 16.65 12.14 0.21 34.27 31.31 9.46 0.71
Malaysia 13.02 15.90 18.11
c
0.44 24.55 28.58 14.10 0.97
New Zealand 7.37 8.49 13.25 0.45 13.52 15.49 12.69 1.01
Pakistan 18.99 22.56 15.80
b
3.82
a
35.67 44.73 20.25
b
3.82
a
Philippines 15.99 14.70 8.82 2.44
b
30.61 28.84 6.13 3.28
a
Shanghai 16.84 15.45 8.98 0.52 31.42 24.88 26.29
b
1.53
Singapore 13.52 16.94 20.20
b
2.10
b
25.01 29.44 15.05
c
2.33
b
Thailand 17.45 19.36 9.87 2.04
b
33.35 35.59 6.27 2.93
a
Taiwan 14.29 14.09 1.39 1.12 26.61 25.67 3.68 1.93
c
Average 14.12 14.59 2.29 0.76 26.17 27.09 2.46 1.43
Total average 14.75 15.32 3.48 0.62 27.12 29.05 5.64 1.21
a
Those test statistics that are statistically signicant at the condence levels of 99%.
b
Those test statistics that are statistically signicant at the condence levels of 95%.
c
Those test statistics that are statistically signicant at the condence levels of 90%.
J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169 1167
It is evident that calculating VaR over a short horizon, followed
by SRTR scaling to convert to longer-term tail risks, is likely to be
inappropriate and misleading, particularly for markets in Eastern
Europe, Central and South America, or the Asia Pacic. Caution is
necessary in applying the SRTR. This investigation offers a new ap-
proach to verifying and recommending the practical applicability
of the SRTR, considering the potential presence of a mixture of con-
founding dynamic dependence, distributional properties and
jumps. This study proposes performing the simple subsample-
based test described here, to complement to a typical variance ra-
tio test in verifying the existence of a synthetic bias based on the
interaction of these underlying effects. The variance ratio test,
w*(h), helps identify the rst-order effect causing the bias, i.e., it
justies the existing serial dependence. Restated, a positive (nega-
tive) w*(h) indicates positive (negative) series dependence. To illus-
trate the practical usefulness of the new procedures, this study rst
considers the 10-day case.
Eleven markets had signicantly positive w*(10) and four had
signicantly negative w*(10). By focusing on the markets exhibiting
persistent returns (a positive w*(10)), their average f(10) was
13.59% which is over twice the average for all markets. However,
the average f(10) was 7.39% among the four mean-reverting mar-
kets (a negative w*(10)). For the 30-day case, 15 markets exhibited
signicantly positive w*(30), and their average f(10) was 18.7%.
Therefore, the empirical results reveal that positive serial depen-
dence indeed causes the SRTR to severely understate the time-
aggregated VaR.
As discussed in Section 3.3, volatility clustering and jump com-
ponents both only slightly distort the approximation performance
of SRTR. Nonetheless, a heavy-tail exerts a greater inuence than
the above two factors. The average kurtosis of 1-day, 10-day, and
30-day returns for all markets are 13.98, 6.80, and 5.44, respec-
tively. Fig. 5 shows the scatter plots of the kurtosis of daily returns
against the corresponding kurtosis of 10-day and 30-day returns
and indicates that almost all kurt[r
t
] are larger than the corre-
sponding kurt[R
k
t
(h)]. Tail heaviness is decreasing through tempo-
ral aggregating daily returns into 10-day or 30-day returns. As
argued previously, VaR (10) or VaR (30) is overestimated owing to
the smoothing of the heavy-tails.
5.3. New ndings after serial dependence correction
To illustrate the practical usefulness of the new procedures, the
main results after implementing the modied-SRTR are summa-
rized in Table 4. By comparing the robust modied-SRTR with
the traditional raw SRTR, the cross-sectional average of the overall
bias, f(10), from the 47 market indices improved from 5.7% to
3.48% and that for the longer horizon, f(30), improved from
13% to 5.64%. Therefore, this robust scaling rule via MVaR in-
duces less bias when estimating the multi-period VaR. In particu-
lar, more evident results indicating the robustness of MVaR could
be found in Table 4. When estimating the 10-day VaR using the
SRTR, there are 10 markets that are signicantly biased; however,
using the modied-SRTR, only ve markets remain signicantly
biased. Similar results are found in the cases for the 30-day VaR.
The number of signicant biased markets is reduced from 12 to
5. In other words, most strong upward or downward biases dis-
closed as signicant in Table 3 are largely due to the failure to
accommodate the serial time dependence of return in the VaR scal-
ing using the raw SRTR. Once such dependence is properly ad-
justed, the bias from the rst-order effect is largely mitigated.
If there remains any signicant bias after adjusting for the time
dependence using the new MVaR, it might be contributed by other
bias factors apart from the time dependence. For instance, if the
remaining bias is negative, this may suggest the existence of a
strong GARCH effect, a non-negligible negative mean or a combina-
tion of both. Similarly, a signicant upward bias may be attribut-
able to the strong inherent infrequent jump components or
heavy-tail phenomenon.
Fig. 6 characterizes the overall bias for 10-day and 30-day VaR
computed from the newly-proposed modied-SRTR, MVaR, across
all markets in the sample. Interestingly, the dependence correction
that makes a notable improvement over the traditional unadjusted
raw SRTR is visually evident in Fig. 6. Obviously the scatters of the
biases are now more centered around zero with bias magnitudes
much smaller than that presented in Fig. 4.
6. Conclusions
Scaling with the SRTR is simple and has been widely employed
in practice, and, even in some instances is required by regulation,
as a tool for approximating longer horizon tail risks in the nancial
industry. The ugly facts based on the real world asset returns make
the optimistic pre-assumptions on which SRTR scaling is built far
from credible, and thus the performance of SRTR scaling is doubt-
ful. This study examines and reconciles different potential bias fac-
tors in nancial return series from the literature to clarify how
biased the SRTR may be by considering alternative return charac-
teristics: including serial dependence, volatility clustering, heavy-
tails, and jumps. By complementing the variance ratio test, this
study proposes a new test that is both intuitive and simple of the
overall validity of the SRTR based on subsampling.
This study nds that serial dependence severely biases the
applicability of SRTR, and that the heavy-tail results in an upwards
bias for the SRTR; By contrast, a non-zero mean in the daily level
gives rise to only mild bias, so that the effect of overlooking jumps
or volatility clustering is less relevant in scaling time-aggregated
VaR. The empirical evidence presented in this study, covering 47
developed and emerging markets included in the MSCI index,
shows that the raw SRTR is a lenient rule, which on average under-
estimates 10-day and 30-day VaR. This implies that nancial insti-
tutions using SRTR will be insufciently prudent and will fail to
resolve their inadequate capital requirements.
On the one hand, for some developed markets, even when the
pre-assumptions are violated, the SRTR scaling yields results that
are generallycorrect, as is showninthe global investigation. This sit-
uation occurs because the underestimation arising from the dy-
namic dependence structure is counterbalanced by the
0 10 20 30 40 50 60
40
30
20
10
0
10
20
30
40
VaR (h) %

(
h
)

%
h=10
h=30
Fig. 6. Scatter Plot of VaR (h) with the corresponding bias f(h) from MVaR. Based on
47 markets, we sketch two scatter plots of VaR (10) and VaR (30) with their
corresponding f(10) and f(30) produced under our newly-proposed modied
scaling MVaR (h) as

h VR(h)
p
VaR(1) VaR(h) respectively, from January 2, 1997
to December 31, 2009.
1168 J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169
overestimation arising from the excess kurtosis and jumps. Hence
the SRTR scaling may serve in its place in assessing the multi-period
VaR. On the other hand, the SRTR scaling of tail risk is likely to be
very inappropriate and misleading, particularly for markets in East-
ern Europe, Central and South America, and the Asia Pacic.
Although its widespread use as a tool for approximating horizon
conversion is understandable, caution is necessary.
As we have shown that the dynamic serial dependence serves as
the rst-order effect that biases the validity of the SRTR among the
other stylized features, the present paper lls this void by develop-
ing a simple and robust scaling rule utilizing the estimated vari-
ance ratio, the modied-SRTR (MVaR (h)), for estimating and
scaling the multi-horizon tail risks. Interestingly, it turns out that
the dependence correction is a notable improvement over the tra-
ditional unadjusted raw SRTR. This study concludes that the use of
certain pretests, as proposed above, is an important step and may
illuminate the applicability of the original SRTR in practical tail risk
approximation. Given the demonstrated performance and their
empirical simplicity, the newly-proposed test as well as the mod-
ied-SRTR approach are likely to appeal to researchers and practi-
tioners alike when estimating longer horizon VaRs.
Acknowledgements
The authors are very grateful to an anonymous referee for his/
her helpful comments and suggestions that lead to substantial
improvements of the paper. Wangs work on this paper was partly
funded by the National Science Council in Taiwan(NSC 98-2410-H-
159-010) and Yeh thank the National Science Council in Taiwan for
nancial support via Grant No. NSC 96-2415-H-008-009. The usual
disclaimer applies.
References
Andersen, T.G., Benzoni, L., Lund, J., 2002. An empirical investigation of continuous-
time equity return models. Journal of Finance 57, 12391284.
Ayadi, O.F., Pyun, C.S., 1994. An application of variance ratio test to the Korean
securities market. Journal of Banking and Finance 18, 643658.
Basel Committee on Banking Supervision, 1996a. Amendment to the capital accord
to incorporate market risk. Bank for International Settlements.
Basel Committee on Banking Supervision, 1996b. Overview of the amendment to
the capital accord to incorporate market risk. Bank for International
Settlements.
Becker, R., Clements, A.E., McClelland, A., 2009. The jump component of S&P 500
volatility and the VIX index. Journal of Banking and Finance 33, 10331038.
Bollerslev, T., Chou, R.Y., Kroner, K.F., 1992. ARCH modeling in nance. Journal of
Econometrics 52, 559.
Bollerslev, T., Engle, R.F., Nelson, D.B., 1994. ARCH model. In: Engle, R.F., McFadden,
D.C. (Eds.), Handbook of Econometrics, vol. 4. Elsevier Science, Amsterdam, pp.
29593038.
Cmara, A., 2009. Two counters of jumps. Journal of Banking and Finance 33, 456
463.
Christoffersen, P.F., Diebold, F.X., 1997. Optimal prediction under asymmetric loss.
Econometric Theory 13, 808817.
Dacorogna, M.M., Mller, U.A., Pictet, O.V., de Vries, C., 2001. Extremal forex returns
in extremely large datasets. Extremes 4, 105127.
Danielsson, J., Zigrand, J.-P., 2006. On time-scaling of risk and the square-root-of-
time rule. Journal of Banking and Finance 30, 27012713.
Diebold, F., Hickman, A., Inoue, A., Schuermann, T, 1997. Converting 1-day volatility
to h-day volatility: scaling by

h
_
is worse than you think. Discussion Paper
Series, No. 97-34, Wharton.
Diebold, F., Hickman, A., Inoue, A., Schuermann, T., 1998. Scale models. Risk 11,
104107.
Drost, F.C., Nijman, T.E., 1993. Temporal aggregation of GARCH processes.
Econometrica 61, 909927.
Engle, R.F., 1982. Autoregressive conditional heteroscedasticity with estimates of
the variance of United Kingdom ination. Econometrica 50, 329.
Eraker, B., Johannes, M.S., Polson, N.G., 2003. The impact of jumps in volatility and
returns. Journal of Finance 58, 12691300.
Fama, E., 1965. The behavior of stock market prices. Journal of Business 38, 34105.
Huang, B.-N., 1985. Do Asian stock market prices follow random walks? Evidence
from the variance ratio test. Applied Financial Economics 5, 251256.
Huang, X., Tauchen, G., 2005. The relative contribution of jumps to total price
variance. Journal of Financial Econometrics 3, 456499.
Jansen, D.W., de Vries, C.G., 1991. On the frequency of large stock returns: putting
booms and busts into perspective. Review of Economics Statistics 73, 1824.
Jorion, P., 2001. Financial Manager Handbook 20012002. John Wiley & Sons, Inc.,
New York.
Lo, A.W., 2002. The statistics of Sharpe ratios. Financial Analysts Journal 58, 3652.
Lo, A.W., MacKinlay, A.C., 1988. Stock market prices do not follow random walks:
evidence from a simple specication test. The Review of Financial Studies 1, 41
66.
Merton, R.C., 1976. Option pricing when the underlying stock returns are
discontinuous. Journal of Financial Economics 5, 125144.
Mller, U.A., Dacorogna, M.M., Olsen, R.B., Pictet, O.V., Schwartz, M., Morgenegg, C.,
1990. Statistical study of foreign exchange rates, empirical evidence of a price
change scaling law, and intraday analysis. Journal of Banking and Finance 14,
11891208.
Pagan, A., 1996. The econometrics of nancial markets. Journal of Empirical Finance
3, 15102.
Pan, J., 2002. The jump-risk premia implicit in options: evidence from an integrated
time-series study. Journal of Financial Economics 63, 350.
Prignon, C., Smith, D.R., 2010a. Diversication and value-at-risk. Journal of Banking
and Finance 34, 5566.
Prignon, C., Smith, D.R., 2010b. The level and quality of value-at-risk disclosure by
commercial banks. Journal of Banking and Finance 34, 362377.
J.-N. Wang et al. / Journal of Banking & Finance 35 (2011) 11581169 1169